Theoretical Study of Conversion and Decay Processes of Excited

May 30, 2017 - Elliot J. TaffetYoann OlivierFrankie LamDavid BeljonneGregory D. Scholes. The Journal of Physical Chemistry Letters 2018 9 (7), 1620-16...
0 downloads 0 Views 2MB Size
Article pubs.acs.org/JPCC

Theoretical Study of Conversion and Decay Processes of Excited Triplet and Singlet States in a Thermally Activated Delayed Fluorescence Molecule Qian Peng,† Di Fan,† Ruihong Duan,† Yuanping Yi,*,† Yingli Niu,† Dong Wang,‡ and Zhigang Shuai*,‡ †

Beijing National Laboratory for Molecular Sciences, CAS Key Laboratory of Organic solids, Institute of Chemistry, Chinese Academy of Sciences, Beijing 100190, China ‡ Key Laboratory of Organic OptoElectronics and Molecular Engineering, Department of Chemistry, Tsinghua University, Beijing 100084, China S Supporting Information *

ABSTRACT: Quantitative understanding of the photophysical processes is essential for developing novel thermally activated delayed fluorescence (TADF) materials. Taking as an example 1,2,3,5-tetrakis(carbazol-9-yl)-4,6-dicyanobenzene, a typical TADF-active molecule, we calculated the interconversion and decay rates of the lowest excited singlet and triplet states at different temperatures as well as the prompt and delayed fluorescence efficiencies at 300 K at the first-principles level. Our results can reproduce well the experimentally available data. It is found that the reverse intersystem crossing rate (kRISC) is sharply increased by 3 orders of magnitude, while the other rates increase slightly or remain unchanged when the temperature rises from 77 to 300 K. Importantly, kRISC reaches up to 1.23 × 106 s−1 and can compete with the radiative and nonradiative decay rates of S1 (1.11 × 107 and 2.37 × 105 s−1) at 300 K, leading to an occurrence of delayed fluorescence. In addition, our calculations indicate that it is the freely rotational motions of the carbazolyl between two cyano groups that are responsible for the interconversion between S1 and T1. The large torsional barriers of other three adjacent carbazolyl groups block the nonradiative decay channels of S1 → S0, leading to strong fluorescence. This work would provide useful insight into the molecular design of high-efficiency TADF emitters. (1CT) and triplet (3CT) states owing to significantly reduced energy gap.2,4 Recently, it was suggested that RISC from 3CT to 1CT involves second order process with nonadiabatic coupling with the local exciton triplet (3LE) states as intermediate state which lies closer to 1CT.7−10 For instance, by using photoinduced absorption measurements combined with quantum dynamics simulations, Etherington et al. observed that the spin−orbit coupling between 1CT and 3CT is mediated by the lowest local exciton triplet (3LE) of donor in 2,8-di(10H-phenothiazin-10-yl)dibenzo[b,d]thiphene-5,5-dioxide.7,8 Through quantum chemical calculations, Chen et al. obtained extremely small direct RISC rate constant for T1 → S1 under Fermi’s golden rule and thus highlighted the important role of nonadiabatic transition at S1−S2 and T1−T2 conical intersection points in the potential energy surfaces (PESs) at the level of TDDFT/BMK/6-31G(d).9 Marian proposed that an energetically close-lying LE state can mediate the triplet-to-

1. INTRODUCTION Recently, the development of thermally activated delayed fluorescence (TADF) materials has become a research hot spot owing to the efficient upconversion of electro-pumped triplet states into singlet for organic light-emitting diodes (OLED).1−3 In the electroluminescence device, electrons and holes are electrically injected from electrodes to form bound states (excitons) in the active layer, with statistically 25% singlet and 75% triplet excitons. Generally, for organic molecules only the singlet excitons fluoresce, rendering a limit of 25% for the internal quantum efficiency. TADF emitters can effectively convert the lowest triplet state (T1) into the lowest singlet state (S1) through reverse intersystem crossing (RISC) when the temperature rises, which largely improves the efficiency of exciton utilization and even makes it reach 100% in principle.4,5 To obtain highly efficient TADF material, it is crucial to understand the ISC mechanism of the interconversion processes of T1 ↔ S1. Some efforts have been made to reveal the mechanism of RISC through experimental measurements and theoretical calculations.1−10 Cyclic intersystem crossing was thought to mostly occur between charge-transfer single © 2017 American Chemical Society

Received: January 22, 2017 Revised: May 22, 2017 Published: May 30, 2017 13448

DOI: 10.1021/acs.jpcc.7b00692 J. Phys. Chem. C 2017, 121, 13448−13456

Article

The Journal of Physical Chemistry C

2. METHODOLOGY 2.1. Optical Spectra. On the basis of the Fermi’s golden rule (FGR) and the Franck−Condon principle, the absorption spectrum σab(ω, T) and the emission spectrum σem(ω, T) read12

singlet upconversion in ACRXTN based on the spin−orbit coupling constants and potential energy profiles calculated at the DFT/MRCI level.10 These works demonstrated that the direct spin−orbit coupling between S1 and T1 is negligible or even forbidden and that the LE states close to T1 are important as an intermediate state for the second-order process for the donor−acceptor (DA)-type TADF emitters. In fact, the second-order process has been already proposed in a case study for anthracene by the authors.11 The experimentally measured ISC rate value from S1 to T2 of anthracene can only be reproduced by adopting the ISC rate formalism based on the second-order perturbation theory combining nonadiabatic coupling and spin orbit coupling.11 In general, the radiative decay rates (∼106−7 s−1) of TADF emitters are smaller than those of traditional fluorophores of ∼108−9 s−1. Therefore, apart from effective RISC process, decreasing the nonradiatve decay rate of S1 → S0 is also important for improving the quantum efficiency of TADF emitters. Thus, it is imperial to quantitatively predict the interconversion rate constants between the singlet and triplet states and the TADF quantum efficiency from first principles. In this work, we chose a typical TADF emitter, 1,2,3,5tetrakis(carbazol-9-yl)-4,6-dicyanobenzene (4CzIPN, see Scheme 1a) as an example because it has a wealth of

σab(ω , T ) =

4π 2ω |μ |2 ∑ Piν |⟨Θfu|Θiν⟩|2 δ(ω − ωfu ,iν) 3ℏc fi u , ν (1)

4ω3 |μ |2 ∑ Piν |⟨Θfu|Θiν⟩|2 δ(ωiν ,fu − ω) 3ℏc 3 fi u , ν

σem(ω , T ) =

(2) ̂

Here Piv = e−βH /Z i represents the Boltzmann distribution function for the initial-state vibronic manifold at the finite ̂ temperature, in which Z i = Tr(e−βH ) is the partition function; Θ denotes the nuclear vibrational wave function; ν and u are the vibrational quantum numbers of the initial and final states, respectively; μfi is the electric transition dipole moment between the two electronic states |Φi⟩ and |Φf⟩, which is independent of the nuclear coordinate under the Franck− Condon approximation. Under the time-dependent TVCF formalism, eqs 1 and 2 can be recast into the time-dependent forms

Scheme 1

FC σabs (ω , T ) =

2πω |μ |2 3ℏc fi

FC σem (ω , T ) =

2ω3 |μ |2 3 fi 3π ℏc

FC (t , T ) d t ∫ ei(ω−ω )t ρabs,0 fi

(3)



FC (t , T ) d t ∫−∞ e−i(ω−ω )t ρem,0 if

(4)

The thermal vibration correlation functions (TVCFs) ρFC abs,0 (t, −1 −iτf Ĥ f −iτiĤ i FC T) and ρabs,0 (t, T) have the same form Z i Tr[e e ]

L

and can be solved analytically by multidimensional Gaussian integrations.13 Here, τi = −iβ − t/ℏ, τf = t/ℏ, β = (kBT)−1 and Ĥ f (Ĥ i) is the harmonic oscillator Hamiltonian of the final (initial) electronic state. 2.2. Radiative and Nonradiative Decay Rate Constants. The radiative decay rate constant is the integration over the whole emission spectrum

a

Chemical structure of 4CzIPN with key labels. b(b) Excited-state decay and conversion processes, including the fluorescent radiative (kSr ) and nonradiative internal conversion (kSnr) from S1 to S0, the intersystem crossing from S1 to T1 (kISC), the reverse intersystem crossing from T1 to S1 (kRISC), and the fluorescent radiative (kTr ) and nonradiative intersystem crossing (kTnr) from T1 to S0

k r(T ) =

photophysical and spectroscopic data in experiments. The interconversion and decay rate constants (kSr , kSnr, kTr , kTnr, kISC, and kRISC in Scheme 1b) of S1 and T1 were quantitatively calculated using the thermal vibration correlation function (TVCF) rate formalism we developed earlier,11 and the prompt and delayed fluorescence efficiencies were obtained correspondingly. It is found that an effective RISC from T1 to S1 occurs without the help of higher triplet and singlet states. The electronic-vibrational coupling analyses demonstrate that the pathways of interconversion between S1 and T1 is mainly ascribed as the free rotational motions of the carbazolyl group between two cyano groups, and the nonradiative dissipation channels are hampered by the large torsional barriers of other three adjacent carbazolyl groups, thus causing strong fluorescence. This work would provide design routes for high-performing TADF materials.

∫ σem(ω , T ) dω

(5)

According to the FGR and the second-order perturbation theory, the general nonradiative decay rate constant can be expressed as kf←i

2π = ℏ

∑ Piv v ,u

Hf′u ,iv +

∑ n,μ

Hf′u , nμ Hn′μ ,iv E iv − Enμ

2

δ(E iv − Efu) (6)

where H′ denotes the interaction between two different Born− Oppenheimer states, consisting of two contributions as follows: BO SO Ĥ ′Ψiv = Ĥ Φi(r; Q) Θv(Q) + Ĥ Φi(r; Q) Θv(Q)

(7)

Here Ĥ BO is the nonadiabatic coupling, and Ĥ SO is the spin− orbit coupling. 13449

DOI: 10.1021/acs.jpcc.7b00692 J. Phys. Chem. C 2017, 121, 13448−13456

Article

The Journal of Physical Chemistry C

The spin−orbit coupling (SOC) matrix elements were treated as perturbation through state interaction,19 and the SOC operator20 results from the spin separation of the parent exact two-component Hamiltonian.21 The SOC calculations were carried out at the TD/B3LYP/6-31G(d) level using the BDF program package.22 The transition dipole moments from the T1 to S0 were computed from the quadratic response function using the Dalton program package.23,24

The intersystem crossing rate constant from initial singlet/ triplet to triplet/singlet states can be recast based on the second-order perturbation theory as (0) (1) (2) k fISC ←i = kf←i + kf←i + kf←i

(8)

where k f(0) ←i ≡ k f(1) ←i

1 |HfiSO|2 2 ℏ

⎡2 ≡ Re⎢ 2 ⎢⎣ ℏ

k f(2) ←i ≡

1 ℏ2



∫−∞ dt eiω t ρfi(0)(t ) if

∑ HfiSO Tif,k ∫



dt eiωif t

−∞

k

∑ Tif,k Tfi,l ∫



−∞

k ,l

(9)

⎤ ⎥ ρfi,(1) ( t ) k ⎥⎦

dt eiωif t ρfi,(2) (t ) kl

3. RESULTS AND DISCUSSION 3.1. Geometric Structures. It is still a challenging issue to choose a suitable exchange-correlation functional to accurately describe the excited-state properties of donor−acceptor-type molecules under the framework of TDDFT,25−29 owing to the following reasons: spurious self-interaction,30 missing derivative discontinuity,31,32 adiabatic approximation in which the memory effects in time-dependent density evolution are neglected,33 orbital relaxation energies and two-electron terms considered insufficiently in the linear response formalism,34 and limitation of single-reference character.35 Many efforts have been devoted to improving the description of electronic transition with charge-transfer (CT) character, including range-separated density functionals with individually tuned parameters36 and optimal HF% in the XC functional under the framework of TDDFT,37 ΔSCF-DFT approach,38−40 and multireference DFT method.35 In addition, TDDFT encounters much difficulty in describing triplet excited states due to the triplet instabilities.41,42 Recently, some benchmarks have been carried out to estimate the energy gap between the first excited singlet and triplet states for the TADF systems.43 To the best of our knowledge, there does not exist any functional to well define both the first excited singlet and triplet states for CT molecules. The molecular geometries in the S0, S1, and T1 states of 4CzIPN in the gas phase were optimized using different density functionals. Figure 1 displays the frontier orbitals and excited-

(10)

(11)

Tif,k(l) is the mixed spin−orbit and nonradiative couplings between two electronic states for the k(l)th normal mode Tif, k(l) =

̂ ∑ (HiSO n ⟨Φn|Pfk|Φf ⟩ / ΔEnf

̂ + HnSO f ⟨Φi |Pnk|Φn⟩ / ΔE in)

n

ρfi(0)(t) is the same as that in spectrum formula, ̂ ̂ ρ f i(,1k ) (t) and ρ f i(, k2l) (t) are Z i−1 Tr[Pfk̂ e−iτf Hf e−iτiHi] ̂ ̂ Z i−1 Tr[Pfk̂ e−iτf Hf Pfl̂ e−iτiHi], respectively.

and and

The internal conversion rate constant between two electronic states with same spin manifold under can be written based on the first-order perturbation theory as14 k fIC ←i ≡

1 R kl ℏ2



∫−∞ dt eiω t ρfiic (t , T ) if

(12)

where Rfi,k ≡ ⟨Φf|P̂ fk|Φi⟩⟨Φi|P̂ fl|Φf⟩ is the nonadiabatic electronic coupling matrix element. ρicfi (t, T) is the same as ρ(2) fi,kl(t) in eq 11. The detailed derivation and solution of equations can be found in our previous work.11,14,15 2.3. Computational Details. The geometry optimizations and frequency calculations were performed for the S0 and T1 states by density functional theory (DFT), and for the S1 state by time-dependent DFT (TDDFT). Six exchange-correlation functionals, namely, B3LYP, mPW1LYP, mPW1B95, BMK, M062X, and M06-HF with different HF exchange percentages in the XC functional of 20, 25, 31, 42, 54, and 100%, and two range-separated hybrid functionals, CAM-B3LYP and ωB97XD, are used combining with the 6-31G(d) basis set. The solvent environments were mimicked by using the polarizable continuum model (PCM) with the equilibrium solvation applied for the geometry optimization and the nonequilibrium solvation for the single-point calculation at the equilibrium geometry.11,16 All the calculations were carried out using the Gaussian 09 package.17 On the basis of the electronic properties, the radiative, internal conversion, and intersystem crossing rate constants were evaluated using the TVCF theory, as implemented in the MOMAP program package.11,15,16,18 The potential energy surface (PES) of S1 was considered by using QS1 = QS0 + D or QS1 = QT1 + D relative to that of S0 or T1, respectively. The vector D is the displacement between S0 and S1 or S0 and T1 geometries. Under the frame of the first-order perturbation theory, the nonadiabatic coupling was treated as the force acting on the atomic nuclei through the transition electric field at the level of TD/B3LYP/6-31G(d).14 The transition electric field was evaluated by TDDFT using the Gaussian 09 package.

Figure 1. Frontier orbitals and transition proportion of HOMO → LUMO for the S1 (blue number) and T1 (red number) states at the optimized S0, S1, and T1 geometries for 4CzIPN in the gas phase.

state transition properties at the S0, S1, and T1 geometries obtained at the B3LYP/6-31G(d) level. Obviously, the highest occupied molecular orbital (HOMO) is mainly localized on the carbazolyl groups while the lowest unoccupied molecular orbital (LUMO) is distributed on the dicyanobenzene moiety. Both the S1 and T1 transitions exhibit the feature of outstanding charge transfer (CT) from carbazolyl groups to dicyanobenzene 13450

DOI: 10.1021/acs.jpcc.7b00692 J. Phys. Chem. C 2017, 121, 13448−13456

Article

The Journal of Physical Chemistry C moiety and very minor local excitation (LE) on central phenyl ring at the optimized S0, S1, and T1 geometries. It is also seen that the electronic density of HOMO becomes more localized to the three neighboring carbazolyl moieties in the optimized S1 state. This type of density localization is always relevant to bond twisting or termed as twist-induced intramoleclular charge transfer (TICT). The torsional angles between the electrondonating and the electron-withdrawing moieties are expected to play a crucial role in determining the CT character. The calculated torsional angles between carbazolyl and dicyanobenzene moieties (θ1, θ2, θ3, and θ4 in Scheme 1a) using different functionals, are presented in Figure 2 (detailed data

These indicate that the amount of CT is sensitive to the choice of xc-functional. The geometric modifications between S1 and S0 play an important role in the photophysical properties because they can largely control the optical spectrum shape and the nonradiative decay rate based on the Franck−Condon principle.45 Table 1 lists the modifications of the key torsional angles and bond lengths from S1 to S0 states at different computational levels. The changes of torsional angles are very susceptible to functionals, and there is not a monotonous variation as a function of the HF fraction in XC functional. B3LYP and mPW1LYP with the smallest HF% give the largest change of ca. 80°. The M062X functional with 54% HF fraction gives the smallest modification of 16°. And the other density functionals show a moderate change of 24° ∼ 45°. These results suggest that the computed photophysical properties would be diverse by using different density functionals. In addition, the largest modifications of bond lengths appear in the carbon−carbon bonds in the central phenyl ring and they are insensitive to density functional. The molecular geometries at the three states were also optimized in toluene at the M062X level, which are almost identical to those in the gas phase (see Table S3 of ESI). This indicates that nonpolar solvent would hardly affect the molecular geometry. By and large, it is of vital importance to choose a suitable functional according to a certain criterion. Stokes shift between the absorption and emission spectra is usually used to assess the performance of an approach to calculate the lowest singlet excited state. Stokes shift is approximately equal to the total geometric reorganization energy at S1 and S0 potential energy surfaces in harmonic oscillator model.46 Thus, large reorganization energy would lead to remarkable Stokes shift and vice versa. Table 1 presents the total reorganization energy (λreorg) at the ground-state potential energy surface at different density functional levels. The contribution of each normal mode to the reorganization energy is given in Figure S1 of ESI. Here the reorganization energy of the ith normal mode (λi) is calculated by λi = Siℏωi = ω2i ΔD2/2 (Si and ωi represent the Huang−Rhys factor and i frequency of the ith normal mode, and ΔDi is the displacement along the ith normal mode coordinate between the equilibrium geometries of two electronic states). As expected, the change trend of λreorg is very similar to the geometric structure modification. The reorganization energy given by M062X is the smallest and the main contributions come from a low-frequency torsional vibration of 35 cm−1 and a high-frequency carbon− carbon stretching vibration of 1627 cm−1. M062X functional predicts the smallest Stokes shift (v = 2λreorg) of 0.56 eV (0.54 eV in toluene), closest to the experimental value of 0.30 eV.2 Table 2 gives the absorption and emission energies (vertical

Figure 2. Density functional dependence of selected geometric parameters (dihedral angles) at the equilibrium geometries of the S0, S1, and T1 states for 4CzIPN in the gas phase.

can be found in Table S1 of Electronic Supplementary Information (ESI)). It is easily seen that among the S0, S1, and T1 states, the geometric structure of S1 is most sensitive to the choice of functionals. The predicted torsional angle increases with the decreasing HF fraction. For instance, θ3 varies from about −51.6° to −89.6° with the decrease of the HF% in XC functional from 100% to 20%. The amount of charge transferred from donor to acceptor are calculated and given in Table S2 of ESI. As expected, the amount of CT is more pronounced with large torsional angles, as observed in many previous systems such as benzylidene malononitrile.44

Table 1. Modifications of the Key Torsional Angles (deg) and Bond Lengths (Å) and the Reorganization Energy (λreorg, eV) for the Transition from S1 to S0 Obtained at Different Density Functional Levels for 4CzIPN

B3LYP mPW1LYP mPW1B95 BMK M062X M06-HF CAMB3LYP ωB97XD

HF%

Δθ1

Δθ2

Δθ3

Δθ4

Σ|Δθ|

LC1−C6

LC3−C4

λreorg

20% 25% 31% 42% 54% 100%

−1.8 −2.3 −0.2 −1.9 −2.8 −5.9 −3.7 −2.6

−26.4 −24.8 −10.8 −23.8 −5.8 −20.0 −17.4 −8.5

−26.6 −26.8 −6.8 −10.7 −1.6 −16.9 −13.3 −4.6

−26.4 −24.8 −11.0 −6.6 −5.9 −20.0 −10.7 −8.5

81.2 78.7 28.8 43.0 16.1 62.8 45.1 24.2

0.044 0.046 0.040 0.047 0.040 0.045 0.041 0.042

0.044 0.046 0.040 0.044 0.040 0.045 0.041 0.042

0.98 1.05 0.45 0.75 0.28 1.14 0.31 0.32

13451

DOI: 10.1021/acs.jpcc.7b00692 J. Phys. Chem. C 2017, 121, 13448−13456

Article

The Journal of Physical Chemistry C

approximations (see detailed derivation in ESI), and the activation energy is hence obtained as ΔGa = (−ΔEif + λreorg)2/ (4λreorg). Herein, ΔEif = Ei − Ef denotes the adiabatic energy difference between the initial T1 and final S1 states. λreorg represents the reorganization energy at the potential energy surface of the S1 state, which could be calculated by using twopoint method based on the excitation energy, namely, λreorg = ES1 (T1‑geometry) − ES1 (S1‑geometry) [ES1 (T1/S1‑geometry) is the S1 energy at the T1/S1 equilibrium geometry]. Figure 3 plots the results of ΔEST, λreorg and ΔGa obtained by different density functionals at the M062X-optimized geometry. The corresponding dependence of ΔEST and λreorg on the HF fraction in the XC functionals is plotted in Figures S2 and S3 of ESI, respectively. The ΔEST and ΔGa increases with the HF fraction in the XC functional while the λreorg decreases. Because of the overestimation of ΔEST and extremely small λreorg, the ΔGa is largely overestimated by M06-HF. The LRC functionals CAMB3LYP and ωB97XD give an unreasonable ΔGa owing to a too large ΔEST of ca. 0.6 eV. Compared with the experiments, B3LYP offers the best result of 0.092 eV, in very good agreement with the measured values of 0.084 eV in toluene and 0.1 eV in thin film.2 Therefore, the B3LYP-calculated excitation energies seem to be most reliable and they would be used for the following rate constant calculations. 3.3. Interconversion Rates and Fluorescence Efficiencies. We first searched the minimal energy path (MEP) of S1 state starting from T1 to S1 equilibrium geometries at level of TD-M062X/6-31G(d) and calculated the potential surfaces of the S1, T1 and T2 states along the MEP by using both M06-2X and B3LYP functionals (see section VIII and Figure S4 in ESI). It is found that S1 is always above T1 and below T2 at the level of the hybrid TD-M062X/B3LYP, and T2 is not considered in the following rate calculations for TADF processes. The calculated average spin−orbit coupling value between S1 and T1 is 0.18 cm−1, which can induce the occurrence of ISC if the energy gap is small enough. Therefore, the thermal vibration correlation function rate formula in the framework of the firstorder perturbation theory can be used to evaluate the interconversion rate constants among the S1, T1 and S0 states. Based on the geometric/electronic properties obtained by the hybrid M062X/B3LYP approach, the calculated rate constants are given in Figure 4a (detailed data seen in Table S5 of ESI), as well as the prompt and delayed fluorescent efficiencies. The displacement and torsion effects of all the normal vibration modes were considered, and the solvent effect was not taken

Table 2. Vertical Energy (eV) at the S0 and S1 Minimum and the Emissive Oscillator Strength at Different Density Functional Levels B3LYP mPW1LYP mPW1B95 BMK M062X M06-HF CAMB3LYP ωB97XD exp. (ref 2)

ΔEv (S0min)

ΔEv (S1min)

Stoke shifts

oscillator strengths

2.49 2.66 2.86 3.09 3.35 4.03 3.36 3.46 2.76

1.70 1.88 2.28 2.41 2.94 2.06 2.76 2.98 2.45

0.79 0.78 0.58 0.68 0.41 1.97 0.60 0.48 0.30

1.67 × 10−5 6.68 × 10−5 0.037 0.024 0.087 0.011 0.024 0.062

excitation energy), and the emissive oscillator strengths, obtained with the different functionals, as well as available experimental values. As expected, M062X give the largest oscillator strength and the smallest Stokes shift due to the smallest distorted angles at S1 minimum and the smallest geometrical modification from S1 to S0, which are in accordance with the experimentally observed strong fluorescence. 2 However, the excited state energies obtained by M062X are much larger than the measured values. The calculated oscillator strengths by B3LYP and mPW1LYP are negligibly small, which is not in accordance with the observed fluorescence. The results from other functionals fall between the two cases. To conclude, the M062X-optimized geometries appear to be most reasonable to estimate the ground-state and excited-state geometries of 4CzIPN. 3.2. Excited-State Energies. RISC depends on the energy gap between S1 and T1 (ΔEST), which is calculated by the M062X functional to be 0.35 and 0.32 eV in the gas phase and toluene solvent, respectively, far larger than the experimental value. Thus, it is necessary to further assess the capabilities of different density functionals to predict the excited-state energies based on the M062X-derived ground-state and excited-state geometries (see Table S4 of ESI). The activation energy is a key factor to control RISC from T1 to S1, and it is reasonable to use it as a criterion parameter to evaluate the performance of density functionals in defining both S1 and T1. In experiments, the activation energy could be estimated from the Arrhenius formula kRISC = A exp(−ΔGa/kBT) in combination with the RISC rates measured at different temperatures. Actually, the Arrhenius formula can be derived from the full-quantum rate formula eq 9 by adopting the short-time and high-temperature

Figure 3. (a) Adiabatic energy differences (ΔEST) and reorganization energy (λreorg); (b) activation energies (ΔGa, the dashed horizontal line denotes the experimental value) of the RISC from T1 to S1 calculated using different density functionals based on the M062X-optimized geometries for 4CzIPN. 13452

DOI: 10.1021/acs.jpcc.7b00692 J. Phys. Chem. C 2017, 121, 13448−13456

Article

The Journal of Physical Chemistry C

Figure 4. (a) The calculated conversion and decay rates (in unit of s−1) of S1 and T1 states and the prompt and delayed fluorescence quantum efficiency (Φp =

k rS S k rS + k nr + kISC



andΦd = ∑m = 1 (ΦISC ΦRISC)m Φp with ΦISC =

kISC S k rS + k nr + kISC

and ΦRISC =

kRISC T k rT + k nr + kRISC

) at 300 K and (b) the

temperature dependence of the interconversion and decay rates from 77 to 300 K for 4CzIPN.

gap at different temperatures in Figure 5. When the energy gap is equal to ΔEST or ΔETS, the rate corresponds to kISC or kRISC,

into account because of the hardly influence on molecular electronic structure property as tested above. In order to check how much the calculated rates depend on the choice of the method, we provided the calculated rates using the same functional (M062X and B3LYP) for geometries and energies for comparison in Table S6 of ESI. Since the singlet−triplet energy gap from M062X is too large, the rate constant of RISC is 6 orders of magnitude smaller than the experiment. The radiative rate constant with B3LYP is 4 orders of magnitude less than the experiment because of tiny oscillator strength and the nonradiative decay rate for S1→ S0 is 2 orders of magnitude larger than the experiment because of the overlarge geometrical modification. The hybrid approach of M062X for nuclear and B3LYP for electronic properties is suitable for the DA-type molecule. Strikingly, the RISC rate constant of T1 → S1 is calculated to be 1.23 × 106 s−1 based on the first-order perturbation which is very close to the experimental data of 2.70 × 106 s−1. The temperature dependence of the interconversion rate constants was examined and shown in Figure 4b (detailed data seen in Table S5 of ESI). The activation energy is fit to be 0.051 eV using the Arrhenius formula on the basis of the calculated RISC rate constants in Figure S5 of ESI, which is very closed to the experimental value and smaller than the value computed from the classical model given in section 3.2. This indicates that the present calculated result is reliable and that the quantum tunneling effect slightly facilitates the occurrence of RISC from T1 to S1. As seen in Figure 4b, kSr hardly changes with temperature, as expected.45 The value of kSnr increases with temperature owing to the participation of more vibrational states. The values of kTr and kTnr are expected to be very small due to weak spin−orbit couplings and large energy gap, which would facilitate the reverse conversion from T1 to S1 owing to thermal equilibrium. More interestingly, the kRISC is substantially increased by 3 orders of magnitude while kISC is almost unchanged when temperature rises from 77 to 300 K. At 300 K, the RISC rate can compete with the other rates. Thus, the delayed fluorescence emerges. The resultant prompt and delayed fluorescence quantum efficiencies are 53.5 and 44.4%, respectively, at 300 K, and the total photoluminescent efficiency is 97.9%, which is very close to the experimental value of 94 ± 2% measured in toluene.2 In order to gain a deep insight into the different behaviors of kISC and kRISC, we plot kISC and kRISC as a function of the energy

Figure 5. Intersystem crossing rate constant as a function of energy gap at different temperatures of 4CzIPN. The points marked with solid stars and circles represent kRISC and kISC, respectively, and the black dotted line is the ΔE = λ.

and the function can be approximated as a parabola of ln(k) = constant −

(−ΔE + λ)2 4λkBT

based on linearly displaced

model and short-time approximation. When ΔE = λ, the rate is almost temperature-independent, and the temperature effect would appear until far away from the point of ΔE = λ. Thus, RISC rate is more sensitive to temperature than ISC rate. It is easier to understand the phenomena from the point of view of energy conservation. Assuming temperature is 0 K, the transition starts from the lowest vibration state of the initial electronic state to a higher vibration state of the final electronic state. For ISC, the energy conservation is naturally favorable to a higher vibration state of T1 state owing to ΔEST > 0. However, for RISC, the energy conservation cannot be satisfied because of ΔETS < 0. Only when the temperature rises to a certain value, the energy conservation is achieved; hence, the RISC can occur. Therefore, RISC rate is more suspicious to temperature than ISC rate. To figure out the structure−property relationship during the energy conversion processes of S1 ↔ T1 and S1 → S0, we analyzed the Huang−Rhys factors as well as the displacement vectors of vibration modes with the largest Huang−Rhys factors (see Figure 6). For the conversion between S1 and T1, 13453

DOI: 10.1021/acs.jpcc.7b00692 J. Phys. Chem. C 2017, 121, 13448−13456

Article

The Journal of Physical Chemistry C

Figure 6. Huang−Rhys factors for energy conversion between S1 and T1 (a) and between S1 and S0 (b), as well as the displacement vectors of two important vibration modes with the largest values of Huang−Rhys factor for 4CzIPN.

many low-frequency vibration modes (